Even though it seems to get the voices’ tones and general sounds right, there is unmistakably something off about them.
Google researchers have developed an AI that can produce minute-long musical compositions from text prompts (AFP)
Washington: In a similar manner to how systems like DALL-E generate images from written prompts, Google researchers have developed an AI that can generate minutes-long musical pieces from text prompts and even transform a whistled or hummed melody into other instruments, according to The Verge, an American technology news website, via TechCrunch.
The model, according to the source, is known as MusicLM, and although you can’t experiment with it on your own, the business has uploaded a number of samples that it created using the model.
Impressive examples are provided. There are five-minute pieces produced from just one or two words like “melodic techno,” as well as 30-second samples that sound like actual songs and are created from paragraph-long descriptions that prescribe a genre, vibe, and even specific instruments.
On the demo site, you can hear examples of the music the model creates when you ask it to create 10-second clips of instruments like the cello or maracas, eight-second clips of a particular genre, music that would fit a prison break, or even the difference between the sounds of a beginner and an experienced pianist. In addition, it offers interpretations of terms like “futuristic club” and “accordion death metal,” according to The Verge.
Even though MusicLM can simulate human vocals and seems to get the pitch and general sound of voices right, there is unmistakably an off-putting element to them.
According to The Verge, artificial intelligence (AI) systems have been credited with creating pop songs, copying Bach more accurately than a human could in the 1990s, and composing music to go with live performances.
ALSO READ: How to use the Fossil Smartwatches App